Abstract: Lack of data and volunteers in underdeveloped areas is the bottleneck that restricts global mapping task. In order to solve this problem, we propose a comprehensive geospatial data crowdsourcing task recommendation method to improve the effectiveness of limited volunteers' contributions. In this method, the research area is divided to several task areas using grids, the spatial preference is computed using triangular kernel and temporal preference is computed using exponential time forgetting rate, and TF-IDF is used to compute the semantic preferences of users. The spatio-temporal-semantics comprehensive preference is calculated using the multiplication rule. The initial task-user recommendation list can be obtained based on the spatio-temporal-semantics comprehensive preference. In order to improve the quality of the contribution data, the user reputationis introduce to our recommendation model, and the latent factor model is used to predict the user's reputation for each task area. The initial recommendation list is reordered according to the user's reputation. In order to verify the effectiveness of the method, we choose Islamabad (the capital of Pakistan) as research area because it is an underdeveloped areas with a certain data foundation. The user and crowdsourced data of Islamabad collected by the OpenStreetMap platform are used as the experiment data. The crowdsourced data is randomly divided into training and test set according to 8:2 ratio. The experimental results show that the proposed method in this paper can not only improve the acceptance rate of recommended tasks, but also can impove the quality of the contributions to some extent.

Key words: crowdsourcing, task recommendation, spatio-temporal-semantics preference, user reputation, OpenStreetMap